GA Based Trajectory Planner for Robot Manipulators Sharing a Common Workspace with Adaptive Population Size

Abstract

This paper presents a strategy to modify, on "the fly", the initial population size of a genetic algorithm based trajectory planner for robot manipulators sharing a common workspace. Firstly, some considerations on the effect of varying the population size of the initial population size of the GA over the performance of the obtained paths are investigated. Finally, a fuzzy unit is implemented to adaptively modify the population size accordingly to the navigation conditions of the manipulator system, such as the magnitude of the potential field (used to model the manipulators as moving obstacles), error to goal, and whether or not the manipulators are moving along a stable trajectory. The considered cases correspond to two redundant dual manipulator systems, consisting of two 3 degrees of freedom (dof) and two 7 DOF planar manipulators.